jsl is a Rust implementation of JSON Schema Language ("JSL"), a portable way to describe and validate the structure of JSON data.

The documentation for this crate focuses on making JSON Schema Language work with Rust. For information on JSON Schema Language in general, refer to the documentation on the JSL homepage.

Validating data

The most common use-case for this crate is checking that some JSON input is really valid against a schema. Here's how you'd achieve that use-case:

use serde_json::json;
use jsl::{Registry, Schema, SerdeSchema, Validator, ValidationError};
use failure::Error;
use std::collections::HashSet;

fn main() -> Result<(), Error> {
let demo_schema_data = r#"
{
"properties": {
"name": { "type": "string" },
"age": { "type": "number" },
"phones": {
"elements": { "type": "string" }
}
}
}
"#;

// The SerdeSchema type is a serde-friendly format for representing
// schemas.
let demo_schema: SerdeSchema = serde_json::from_str(demo_schema_data)?;

// The Schema type is a higher-level format that does more validity
// checks.
let demo_schema = Schema::from_serde(demo_schema).unwrap();

// A registry is a bundle of schemas that can cross-reference one
// another. When you add a SerdeSchema to a Registry, the Registry will
// return the URIs of all schemas still missing from the Registry.
let mut registry = Registry::new();
let missing_uris = registry.register(demo_schema)?;

// Our schema doesn't use references, so we're not expecting any
// dangling references to other schemas.
assert!(missing_uris.is_empty());

// Once you've registered all your schemas, you can efficiently begin
// processing as many inputs as desired.
let validator = Validator::new(&registry);
let input_ok = json!({
"name": "John Doe",
"age": 43,
"phones": [
"+44 1234567",
"+44 2345678"
]
});

let validation_errors_ok = validator.validate(&input_ok)?;
assert!(validation_errors_ok.is_empty());

let input_bad = json!({
"age": "43",
"phones": [
"+44 1234567",
442345678
]
});

// Each ValidationError holds paths to the bad part of the input, as
// well as the part of the schema which rejected it.
//
// For testing purposes, we'll sort the errors so that their order is
// predictable.
let mut validation_errors_bad = validator.validate(&input_bad)?;
validation_errors_bad.sort_by_key(|err| err.instance_path().to_string());
assert_eq!(validation_errors_bad.len(), 3);

// "name" is required
assert_eq!(validation_errors_bad[0].instance_path().to_string(), "");
assert_eq!(validation_errors_bad[0].schema_path().to_string(), "/properties/name");

// "age" has the wrong type
assert_eq!(validation_errors_bad[1].instance_path().to_string(), "/age");
assert_eq!(validation_errors_bad[1].schema_path().to_string(), "/properties/age/type");

// "phones[1]" has the wrong type
assert_eq!(validation_errors_bad[2].instance_path().to_string(), "/phones/1");
assert_eq!(validation_errors_bad[2].schema_path().to_string(), "/properties/phones/elements/type");

Ok(())
}

The ValidationError type that Validator::validate produces contains two json_pointer::JsonPointers. These paths within these validation errors are standardized, and should be understood by any implementation of JSL, not just this crate. All compliant implementations of JSL produce the same paths, making them platform-independent.

Writing tooling on top of JSL

JSL was designed with the same principles that make JSON so useful: it's easy to implement, and even easier to build on top of. If you're building custom tooling on top of JSL, such as UI, documentation, or code generation, this crate provides a Schema type for that purpose.

See the docs for Schema for more.